Thiazolidinediones (TZD) are a relatively new class of insulin-sensitizing agents used to treat type 2 diabetes mellitus (T2DM) and have also been shown to reduce risk for, or even prevent, T2DM in at-risk individuals. However, 30-40% of subjects do not respond to TZD therapy. TZDs are agonists for peroxisome proliferator-activated receptor-g2 (PPARG2). We hypothesize that variation in the gene encoding for PPARG2 may mediate response to TZDs. We also hypothesize that variants in genes encoding for proteins involved in the metabolism of TZDs or in the TZD-stimulated pathway may also contribute to response to drug. We propose the following series of studies to address these hypotheses. First, we propose a three- month open-label pioglitazone (PIO) trial in Latinas with previous gestational diabetes (GDM) to increase the sample size of our existing data. Women will be placed on PIO (45 mg/d) for three months and body composition, insulin sensitivity, and b cell function will be assessed at baseline and 3-months. Lack of response to PIO will be determined by a non-significant improvement in insulin sensitivity. This trial will provide additional data to test association between genetic variants and TZD response, and TZD-induced changes in metabolic phenotypes. Second, we propose to genotype genetic variants in PPARG shown to be associated with response to troglitazone and to screen candidate genes;three cytochrome P-450 genes (CYP2C8, CYP2C9, and CYP3A4), which are involved in PIO metabolism;retinoid X receptor-a (RXRA) a critical co-factor for PPARG;and peroxisome proliferator-activated receptor-g coactivatoMa (PPARGC1A) a critical regulator of the PPARG-stimulated pathway. Third, we propose specific genetic analyses to test variants genotyped in the second aim for association with response to PIO and PlO-induced changes in phenotypes. We propose methods to control for potential population stratification and multiple comparisons. We also propose exploratory analyses to examine the effect of multiple genetic variants (within genes and between genes) on response to PIO. Our long-term goal is to understand the genetic architecture of TZD response and to develop approaches to predict who will or will not respond to TZDs. This will help clinicians to avoid interventions that have a low probability of success in a given patient with T2DM.